#15379: Don't reject user-written instances of KnownNat and friends in hsig files -------------------------------------+------------------------------------- Reporter: ezyang | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: 8.6.1 Component: Compiler | Version: 8.4.3 Resolution: | Keywords: backpack Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by ppk): Resolving the overlapping instance problem is more tricky than I thought. I am adding additional notes here so that some one can better guide the process. This is my current understanding on the issues The `KnownNat` and `KnownSymbol` instance witness and dictionary generating is done by ghc itself in the file `compiler/typecheck/ClsInst.hs`. However this is problematic in the backpack setting because of the following Let us consider the following signature fragment {{{#!haskell signature Abstract where data Foo :: Nat instance KnownNat Foo }}} Suppose this abstract signature is included in a module `Util` {{{#!haskell module Util where import Abstract printFoo :: IO () printFoo = do print $ natVal (Proxy :: Proxy Foo) }}} Notice that the `natVal` function will need a dictionary of `KnownNat Foo` at this point and will need to get it from a concrete implementation of the kind {{{#!haskell module Concrete where type Foo = 42 }}} It is not clear to me how this can be generated "inside the compiler" like what happens in `ClsInst.hs`. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/15379#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler